[ATM] WAS: RE: ATM Digest, Vol 30, Issue 21; IS: FEA models of test stand induced deflections

["Scott Milligan" <starzkey@charter.net>]

Folks:

I own a licensed copy of the ALGOR FEA program, and have learned just enough
about using it to generate useful models of mirror self-weight deflection
under various support schemes.  I would be willing to run a few test cases
to see if these models help corroborate what has been observed on the test
stand.  

The ALGOR model will give the surface displacement values at the nodes of
the mesh; to make these results truly useful, however, I suspect one wants
to fit these data to a reasonable set of Zernike polynomials, and do the
usual subtraction of piston, tilt, & defocus before reaching conclusions
regarding the suitability of a given mounting scheme.  I haven't myself
gotten around to writing such a routine (I'm sure that MathCad, R, or
similar package could do it), but that is also an area where others in this
group are strong.  

If a routine that resamples the (typically) irregular spacing of the FEA
nodes onto a regular grid were to become available, then the raw
displacement data can be input to a Zemax model, and the usual suite of
image quality analyses then become available.  Zemax can also accept Zernike
input, but resampling the raw data to a uniform grid should, I expect,
improve the accuracy of the Zernike fit as well.

If we go forward with this, what I would need to construct successful FEA
models would be something like the following:

1. Mirror substrate material (Pyrex, Silica, etc.)
2. Diameter, ROC, and thickness.
3. Support scheme description (i.e. 2 pegs @ 90 degrees, stud through a
central perforation, Sling, etc).
4. An initial assumption describing how the support structure actually
establishes contact (i.e. single point in COG plane, line contact, etc.)
5. Orientation of gravity vector wrt to surface normal at mirror vertex.
6. If gravity vector is oriented other than orthogonal to surface normal,
then some description of the backside support is necessary as well (i.e. #
of support points and their locations).

Models with simple geometry (i.e. revolved solids) are considerably less
time consuming to set up and debug than are models containing intricate
features such as lightweighting ribs, socket inserts, etc.  So I suggest we
start with something simple.  An early lesson learned about FEA is that
simple, "known" cases provide a necessary sanity check on the assumptions
used to create the model.

I have successfully executed several models using peg mount and center
support mounts, with what I feel is good qualitative agreement to my own
interferometric results, but have not yet tried to model anything like a
sling.  So if rapid progress counts for anything, I suggest not starting
with a sling mount.

Cheers,


Scott Milligan
 
 

-----Original Message-----
From: atm-bounces@atmlist.net [mailto:atm-bounces@atmlist.net] On Behalf Of
Stephen Koehler
Sent: Friday, June 16, 2006 5:17 PM
To: Wolfgang Rohr
Cc: atm@atmlist.net
Subject: Re: [ATM] ATM Digest, Vol 30, Issue 21

Wolfgang,

> you can help me:
> In Germany we had a mirror testing project. 16 persons were testing 
> one astigmatic mirror and the main problem was the support of that 
> astigmatic mirror. The problem is, how we cause any kind of 
> astigmatism by the support. You can subtract the induced astigmatism 
> if you can calculate the amount of it. And therefore we need a program 
> similar like Plop but to calculate the lateral pressure on a mirror. 
> Is there anybody who can help?

This seems to directly pertain to my recent post to the interferometry
group, where I describe a new (to me) algorithm for separating mirror from
test stand aberrations from Zernike coefficients from the mirror at various
angles of rotation.  My current algorithm works with an arbitrary amount of
rotational data, but can use as few as two rotations.  The rotation angles
are arbitrary, although some combinations are not very useful (e.g.,
rotation of 0 degrees or 180 degrees).

I think it would difficult to calculate test stand aberrations from
simulation in a reliable way.  It's much better just to measure it.
--
Steve Koehler
_______________________________________________
ATM mailing list http://www.atmlist.net/
_______________________________________________
ATM mailing list http://www.atmlist.net/